Title: Physics-based Modeling and Control of Homogeneous Charge Compression Ignition (HCCI) Engines
1Physics-based Modeling and Control of Homogeneous
Charge Compression Ignition (HCCI) Engines
- Gregory M. Shaver
- Dynamic Design Lab
- May 6th, 2005
- Department of Mechanical Engineering
- Stanford University
2Outline
- What is residual-affected HCCI? What are its
benefits? - Hurdles to practically implementing HCCI
- Lack of combustion trigger
- Cyclic coupling
- Dynamic modeling of HCCI
- Making HCCI practical with feedback control
- Conclusions and future work
3What is Residual-Affected HCCI?
- Residual-Affected Homogeneous Charge Compression
Ignition - Advanced combustion strategy for piston engines
- Combustion due to uniform auto-ignition using
compression alone - Hot exhaust gases reinducted using Variable Valve
Actuation (VVA) - Main benefits
- Increased efficiency compared to SI
- Modest compression ratios
- Drastic reduction in NOx emissions (i.e. smog)
4HCCI with Variable Valve Actuation
5HCCI with Variable Valve Actuation
- Reactants (fuel air) previously exhausted
gases (residual) inducted
6HCCI with Variable Valve Actuation
- Reactants (fuel air) previously exhausted
gases (residual) inducted - Compression of mixture
7HCCI with Variable Valve Actuation
- Reactants (fuel air) previously exhausted
gases (residual) inducted - Compression of mixture causes auto-ignition
- uniform, fast uncontrolled
8HCCI with Variable Valve Actuation
- Reactants (fuel air) previously exhausted
gases (residual) inducted - Compression of mixture causes auto-ignition
- uniform, fast uncontrolled
- Useful work from expansion
9HCCI with Variable Valve Actuation
- Reactants (fuel air) previously exhausted
gases (residual) inducted - Compression of mixture causes auto-ignition
- uniform, fast uncontrolled
- Useful work from expansion
- Hot combustion products exhausted
10HCCI with Variable Valve Actuation
- Reactants (fuel air) previously exhausted
gases (residual) inducted - Compression of mixture causes auto-ignition
- uniform, fast uncontrolled
- Useful work from expansion
- Hot combustion products exhausted, portion
reinducted
11HCCI with Variable Valve Actuation
- Valve motions from VVA determine
- inducted gas composition
- amount of compression
12HCCI with Variable Valve Actuation
13HCCI with Variable Valve Actuation
- Sudden rise in pressure combustion
initiation
14HCCI with Variable Valve Actuation
- Sudden rise in pressure combustion
initiation - Work output
15HCCI with VVA -Challenges
- Goal achieve desired combustion timing work
output - Challenges
- No direct initiator of combustion
- Cycle-to-cycle coupling through exhaust gas
- Significantly complicate transient load operation
16HCCI with VVA -Challenges
- Goal achieve desired combustion timing work
output - Challenges
- No direct initiator of combustion
- Cycle-to-cycle coupling through exhaust gas
- Significantly complicate transient load operation
- To date HCCI impractical!!
17Research Goals
- Make HCCI practical through closed-loop control
- Stabilize process control work output
- Modeling Objective Simple physical models that
capture behavior most relevant for control - Cyclic coupling
- Combustion timing
- In-cylinder pressure evolution (work output)
- Control Objective - Control of
- Combustion timing make combustion sure happens!
- Work output the key output of the engine
- efficiency reduced emissions come as result of
process
18Previous Work Simulation Modeling
- Ogink and Golovitchev 2002, Babajimopoulos et al.
2002 - Multi-zone modeling of HCCI
- Kong et al. 2002
- Multi-dimensional CFD models using detailed
chemistry - Many others
- Complex flow and chemical kinetics models
- Capture general steady state behavior
- Ignore cycle-to-cycle coupling
- Exhibit long run times - 12 hours per engine
cycle
19Contributions Simulation Modeling
- Developed a simulation model of residual-affected
HCCI that - Captures the cyclic coupling
- Predicts behavior during steady state
transients - Captures ignition via kinetics with a simple,
intuitive model - runtimes 15 seconds per engine cycle (amenable
to use as a control testbed)
20Previous Work - Control
- Agrell et al. 2003, Haraldsson et al. 2003,
Bengtsson et al. 2004, Olsson et al. 2001,
Matthews et al. 2005, others - Various approaches to control combustion timing
or work output - In all cases controller hand-tuned or
synthesized from black-box models
21Contributions - Control
- Physics-based control model of HCCI
- The first physics-based approach to control of
HCCI - Generalizable
- Enables use of control engineering tools
- Theoretical control design
- Stability analysis
- Control strategies for
- Combustion timing
- Peak pressure or work output
22Outline - Modeling Strategies
- Simulation model
- Gain some intuition of the process
- What are key features?
- What are relevant control inputs outputs?
- Control model
- Need a slightly simpler physical description for
synthesis - The launching point for developing control
strategies - ..making HCCI practical!!
23Experimental Apparatus
- Single cylinder engine
- With VVA
- Fuel used Propane
- Compression ratio
- Variable 13-15.5
- Engine speed
- Fixed 1800 rpm
- In-cylinder pressure transducer
- Combustion timing
- Peak pressure
- Work output
24HCCI Simulation Model
- 1st law analysis of cylinder and exhaust manifold
25HCCI Simulation Model
- 1st law analysis of cylinder and exhaust manifold
- Steady state 1D compressible flow relations
26HCCI Simulation Model
- 1st law analysis of cylinder and exhaust manifold
- Steady state 1D compressible flow relations
- Heat transfer
- In-cylinder (modified Woschni)
- Ref Chang et al. 2004
- Exhaust manifold
27HCCI Simulation Model
- 1st law analysis of cylinder and exhaust manifold
- Steady state 1D compressible flow relations
- Heat transfer
- In-cylinder (modified Woschni)
- Ref Chang et al. 2004
- Exhaust manifold
- Combustion model
- Wiebe function
- What do we use as a combustion trigger?
28HCCI Simulation Model
- 1st law analysis of cylinder and exhaust manifold
- Steady state 1D compressible flow relations
- Heat transfer
- In-cylinder (modified Woschni)
- Ref Chang et al. 2004
- Exhaust manifold
- Combustion model
- Wiebe function
- What do we use as a combustion trigger?
- Resulting Model 9 nonlinear ODEs
29Temperature Threshold
- Assume HCCI occurs at a threshold temperature
- A fit at one temperature
30Temperature Threshold
- Assume HCCI occurs at a threshold temperature
- Fit at one temperature doesnt hold at others!
Increasing residual
31What Happened?
- Simulation model earlier timing for increasing
residual - More residual means mixture temperature
- Higher temperature leads to early timing
- Experiments show more constant timing
- Is some physical effect missing?
- Yes! Concentration of reactants
- More residual means lower reactant concentration
32Integrated Arrhenius Rate Equation
- Simple model for start of combustion
- Integrated Arrhenius rate
- Constant threshold,
- a, b and Ea from published experiments
- Contributions from temperature reactant
concentration captured
33Integrated Arrhenius Rate
- Set threshold at one operating point
34Integrated Arrhenius Rate
- Set threshold at one operating point and
pressure, timing work output at all points is
captured
Increasing residual
35Integrated Arrhenius Rate
- Note can vary composition without much change in
timing -
Increasing residual
36Simulation Model Can it be extended?
- Steady state behavior with propane captured
- What about transients?
- Changes in load
- Can the model capture these?
37Simulation Model Transients
- 1st operating point has higher steady state
temperature than 2nd - The elevated exhaust temperature advances
combustion process during transition - As exhaust temperature decreases, behavior
reaches new steady state
Experiment
38Simulation Model Transients
Experiment
- Simple model captures the coupling and ignition
behavior during transition
Simulation
39Results from Simulation modeling
- Aspects most relevant for control captured with
simple simulation model - Cyclic coupling combustion timing
- In-cylinder pressure evolution
- Approach can handle
- Steady-state behavior
- Transients
- A valuable virtual testbed for control
40Motivation for Control Model
- Simulation model has a lot of benefits
- Still, too complex for synthesizing control
strategies - Motivates a simpler dynamic model
- Enabled through additional physical assumptions
- Discretizing the process (induction, compression,
etc.) - Linking processes
41Control Model Assumptions
- Assumptions
- Induction atmospheric pressure
- Isentropic compression expansion
- HCCI is fast constant volume combustion
- In-cylinder heat transfer of combustion energy
42Control Model Assumptions
43A Simple Control Model
44A Simple Control Model
- Step through process to develop model of dynamics
45A Simple Control Model
- Step through process to develop model of dynamics
dynamics
46Peak Pressure Dynamics
- The peak pressure dynamics takes the form
- Fairly complex nonlinear dynamic model
47Peak Pressure Dynamics
- The peak pressure dynamics takes the form
- Fairly complex nonlinear dynamic model
- Can see dependence on
- Control inputs
48Peak Pressure Dynamics
- The peak pressure dynamics takes the form
- Fairly complex nonlinear dynamic model
- Can see dependence on
- Control inputs
- Cyclic coupling
49Peak Pressure Dynamics
- The peak pressure dynamics takes the form
- Fairly complex nonlinear dynamic model
- Can see dependence on
- Control inputs
- Cyclic coupling
- Combustion timing
50Peak Pressure Dynamics
- The peak pressure dynamics takes the form
- Fairly complex nonlinear dynamic model
- Can see dependence on
- Control inputs
- Cyclic coupling
- Combustion timing
- How do we model initiation of combustion, qcomb?
51Combustion Timing Dynamics
- Recall the integrated Arrhenius rate model
52Combustion Timing Dynamics
- Recall the integrated Arrhenius rate model
- Integrand takes on largest value at TDC
53Combustion Timing Dynamics
- Recall the integrated Arrhenius rate model
- Integrand takes on largest value at TDC
- Simplify begin integration at TDC with values at
TDC
54Combustion Timing Dynamics
- Recall the integrated Arrhenius rate model
- Integrand takes on largest value at TDC
- Simplify begin integration at TDC with values at
TDC
55Combustion Timing Dynamics
- Recall the integrated Arrhenius rate model
- Integrand takes on largest value at TDC
- Simplify begin integration at TDC with values at
TDC
56Combustion Timing Dynamics
- Recall the integrated Arrhenius rate model
- Integrand takes on largest value at TDC
- Simplify begin integration at TDC with values at
TDC - Algebraic expression exists for each variable
57Combustion Timing Dynamics
- Recall the integrated Arrhenius rate model
- Integrand takes on largest value at TDC
- Simplify begin integration at TDC with values at
TDC - Algebraic expression exists for each variable
58Control Model
- Peak pressure and combustion timing dynamics
together give - A nonlinear 2-state, dynamic, discrete system
model
59Control Model Validation
- Control model captures
- Steady state transient
- Peak pressure
- Combustion Timing
- Captures
- Cyclic coupling
- Ignition via kinetics
60Control Modeling Summary
- HCCI is difficult to control
- Cyclic coupling
- No direct combustion trigger
- Control model captures these phenomena!!
- Simple model tells us how dynamics are affected
by control inputs - Is a launching point for
- Synthesizing control strategies
- Assessing system stability
- Generalizable
61Outline of Control Implementations
- From control model
- Peak-pressure control at constant combustion
timing - Work output control at constant combustion timing
- Simultaneous peak pressure and combustion timing
control - Many other approaches possible
62Peak Pressure Control w/ Constant timing
- Fix final valve closure
- Vary composition to control peak pressure
- A static approach to controlling timing
- A large number of control approaches can be
utilized
63Peak Pressure Control w/ Constant timing Linear
Controller Synthesis
- A common control approach is to linearize the
system model - Linearizing about an operating point
yields - Simple linear control laws can be synthesized
where
64Peak Pressure Control w/ Constant timing
- In closed-loop
- Controller synthesized from linearized version of
model - Is controller stable in closed-loop with
nonlinear model?
65Peak Pressure Control w/ Constant timing
Nonlinear Stability Analysis
- Using
- Lyapunov stability theory
- Convex optimization
- Shows
- Simple linear controller stabilizes entire
operating regime
66Peak Pressure Control w/ Constant timing
Experimental Implementation
- Accurate control of peak pressure
- Mean tracking
- Fluctuation reduction
- Increases robustness
- Little change in phase
- What about direct control of work output (IMEP)?
67Experimental Work Output Control
- Rapid mean tracking fluctuation reduction
- We can control work output, while keeping timing
roughly constant
68Experimental Work Output Control
- Positive and negative load transients
- What about simultaneous control of combustion
timing and work output?
69Combustion Timing Work Output Control
- Add other control input final valve closure
- Significant control knob for combustion timing
- Simple approach
- Separate linear controllers for peak pressure and
timing
70Decoupled Peak Pressure and Phase Control
- Maintain cycle-to-cycle peak pressure controller,
vary phase more slowly
71Experiments with Decoupled Control
- Approach works
- Simultaneous control of
- timing and peak pressure
72Comments on Control Experiments
- Simple physics-based controllers works well
- Implementation is straightforward
- Mean tracking fluctuation reduction of
- peak pressure
- work output
- Combustion timing fairly constant
- Independent control of peak pressure combustion
timing - Many others possible
73Conclusion
- HCCI has a promising future as a cleaner, more
efficient strategy - Hurdle controlling the process
- No combustion initiator cycle-to-cycle coupling
- The good news HCCI is amenable to model-based
control - Key behaviors captured in both simulation and
control models - Simulation control models capture
- Steady-state
- Transients
- Physics-based control of
- Peak pressure
- Work output
- Combustion timing
74Future Work
- Study different control approaches
- Control of multi-cylinder HCCI engines
- Results to date with single-cylinder engines
- Cylinder-to-cylinder dynamics now play a key
role! - Change the world!
75Acknowledgments
- Chris Gerdes
- The Dynamic Design Lab
- Partners in crime
- Matt Roelle Nikhil Ravi
- A great sponsor Robert Bosch Corporation
- Jean-Pierre Hathout, Jasim Ahmed, Aleks Kojic
Sungbae Park - The defense Committee
- Chris Edwards, Sanjay Lall, Matt Franchek Steve
Rock - Stanford University